REDUCED FREQUENCY-SELECTIVITY AND THE PRESERVATION OF SPECTRAL CONTRAST IN NOISE

Reduced frequency selectivity associated with sensorineural hearing lo ss may pose particular problems for hearing-impaired listeners in nois y environments. In these situations, broader-than-normal auditory filt ers may affect the perception of speech by reducing the contrast betwe en spectral peaks and valleys in at least two ways. First, the peaks a nd valleys in the internal representation of the speech spectrum becom e smeared, resulting in less precise frequency analysis. Second, there may be a reduction in the signal-to-noise ratio (S/N) at the output o f each auditory filter. In order to examine the relationship between f requency selectivity and identification of speechlike stimuli in noise , hearing-impaired and normal-hearing listeners were trained to assign vowel labels to four harmonic complexes which differed in the frequen cy locations of four elevated (''peak'') harmonics. Peak harmonics wer e chosen to approximate first- and second-formant frequencies in four English vowels. Listeners were then tested to determine the spectral c ontrast necessary between peak and background components in order to m aintain identification accuracy in the presence of various levels of b roadband noise. Results indicated that for these stimuli, normal-heari ng listeners required about 1 dB of additional spectral contrast for e very doubling of the intensity of noise. The required increase in spec tral contrast was generally greater for listeners with broader-than-no rmal auditory filters at 2000 Hz. This finding suggests indirectly tha t in the internal representations of speech sounds embedded in noise, the signal-to-noise ratio for listeners with abnormal frequency select ivity is poorer than for listeners with normal frequency selectivity. A poorer-than-normal internal S/N may be one factor underlying the com mon observation that noise often is more degrading to speech understan ding by hearing-impaired listeners than by normal-hearing listeners.